L09 Themes

Data Visualization (STAT 302)

Overview

The goal of this lab is to play around with the theme options in ggplot2.

Datasets

We’ll be using the cdc.txt and NU_admission_data.csv datasets.

# load package(s)
library(tidyverse)
library(janitor)
library(ggthemes)
library(patchwork)
library(showtext)

# read in the cdc dataset
cdc <- read_delim(file = "data/cdc.txt", delim = "|") |>
  mutate(
    genhlth = factor(
      genhlth,
      levels = c("excellent", "very good", "good", "fair", "poor"),
      labels = c("Excellent", "Very Good", "Good", "Fair", "Poor")
      )
    )

# read in NU admission data
admin_dat <- read.csv("data/NU_admission_data.csv") %>% clean_names()


# set seed
set.seed(2468)

# selecting a random subset of size 100
cdc_small <- cdc |> 
  slice_sample(n = 100)

Exercise 1

Use the cdc_small dataset to explore several pre-set ggthemes. The code below constructs the familiar scatterplot of weight by height and stores it in plot_01. Display plot_01 to observe the default theme. Explore/apply, and display at least 7 other pre-set themes from the ggplot2 or ggthemes package. Don’t worry about making adjustments to the figures under the new themes. Just get a sense of what the themes are doing to the original figure plot_01.

There should be at least 8 plots for this task, plot_01 is pictured below. Use patchwork or cowplot in combination with R yaml chunk options fig-height and fig-width (out-width and fig-align may be useful as well) to setup the 8 plots together in a user friendly arrangement.

# plot
plot_01 <- ggplot(
  data = cdc_small,
  aes(x = height, y = weight)
) +
  geom_point(size = 3, aes(shape = genhlth, color = genhlth)) +
  scale_y_continuous(
    name = "Weight in Pounds",
    limits = c(100, 275),
    breaks = seq(100, 275, 25),
    trans = "log10",
    labels = scales::label_number(
      accuracy = 1,
      suffix = " lbs"
    )
  ) +
  scale_x_continuous(
    name = "Height in Inches",
    limits = c(60, 80),
    breaks = seq(60, 80, 5),
    labels = scales::label_number(accuracy = 1, suffix = " in")
  ) +
  scale_shape_manual(
    name = "Health?",
    labels = c(
      "Excellent", "Very Good",
      "Good", "Fair", "Poor"
    ),
    values = c(17, 19, 15, 9, 4)
  ) +
  scale_color_brewer(
    name = "Health?",
    labels = c(
      "Excellent", "Very Good",
      "Good", "Fair", "Poor"
    ),
    palette = "Set1"
  ) +
  theme(
    legend.position = c(1, 0),
    legend.justification = c(1, 0)
  ) +
  labs(title = "CDC BRFSS: Weight by Height")

Solution

plot_02 <- plot_01 + theme_dark()
plot_03 <- plot_01 + theme_void()
plot_04 <- plot_01 + theme_solarized()
plot_05 <- plot_01 +theme_excel()
plot_06 <- plot_01 + theme_tufte()
plot_07 <- plot_01 + theme(panel.grid.major = element_line(linewidth = 2))
plot_08 <- plot_01 + theme(panel.background = element_rect(fill = "linen"))

plot_01
plot_02
plot_03
plot_04
plot_05
plot_06
plot_07
plot_08

Important

Which theme or themes do you particularly like? Why?

Solution

I like theme_solarized() because it is something different and isn’t the stark white background that we are used to. Similarly, I like when the panel_background within the theme is set to linen - it is a nice, non-distracting background that highlights the points well. On a less distracting plot, setting the major gridlines to be thick, white lines might have a nice effect, but not on this plot.

Exercise 2

Using plot_01 from Exercise 1 and the theme() function, attempt to construct the ugliest plot possible (example pictured below). Be creative! It should NOT look exactly like the example. Since the goal is to understand a variety of adjustments, you should use a minimum of 10 different manual adjustments within theme().

Solution

plot_01 + 
  theme(plot.background = element_rect(fill = "blue2", colour = NA),
        legend.background = element_rect(
          fill = "darkolivegreen1", 
          colour = "grey50", 
          linewidth = 1),
        panel.grid.major = element_line(linewidth = 2, color = "brown"),
        panel.background = element_rect(fill = "darkslateblue"),
        plot.title = element_text(face = "bold", colour = "red"),
        axis.title.x = element_text(color = "burlywood", size = 14),
        axis.title.y = element_text(color = "azure4", size = 14),
        axis.text.x = element_text(angle = 90, hjust = 0.5, color = "limegreen", size = 20),
        axis.text.y = element_text(face = "italic", color = "magenta", size = 18),
        legend.text = element_text(face = "bold", color = "orange", size = 25)
        )

Exercise 3

We will be making use of your code from Exercise 3 on L07 Layers. Using the NU_admission_data.csv you created two separate plots derived from the single plot depicted in undergraduate-admissions-statistics.pdf. Style these plots so they follow a “Northwestern” theme. You are welcome to display the plots separately OR design a layout that displays both together (likely one stacked above the other).

Check out the following webpages to help create your Northwestern theme:

• Visual Identity
• Color Palettes
• Fonts & Typography — Need to use substitute fonts

Additional requirement:

Use a free non-standard font from google for the title. Pick one that looks similar to a Northwestern font.

Note

I find this blog post to be extremely useful for adding fonts. Important packages for using non-standard fonts are showtext, extrafont, extrafontdb, and sysfonts. The last 3 generally just need to be installed (not loaded per session).

Solution

font_add("gothic_font", "CloisterBlack.ttf")
font_add("Georgia", regular = "/Library/Fonts/Georgia Italic.ttf")


showtext_auto()
bar_dat <- admin_dat %>%
  filter(year >= 2002) %>%
  mutate(
    cat_a = applications - admitted_students,
    cat_b = admitted_students - matriculants,
    cat_c = matriculants
  ) %>%
  select(year, contains("cat_")) %>%
  pivot_longer(
    cols = -year,
    names_to = "category",
    values_to = "value"
  )

bar_labels <- admin_dat %>%
  filter(year >= 2002) %>%
  select(-contains("rate")) %>%
  pivot_longer(
    cols = -year,
    names_to = "category",
    values_to = "value"
  ) %>%
  mutate(
    col_label = prettyNum(value, big.mark = ",")
  )

plot_1 <- ggplot(data = bar_dat, aes(factor(year), value)) +
  geom_col(aes(fill = category), width = 0.6, alpha = 0.7) +
  theme_minimal() +
  geom_text(
    data = bar_labels,
    aes(label = col_label),
    size = 2.5,
    color = "black",
    fontface = "bold"
  ) +
  labs(
    x = "Entering Year",
    y = "Applications",
    title = "Northwestern University \nUndergraduate Admissions 2001-2021"
  ) +
  theme(
    plot.title = element_text(family = "gothic_font", size = 20, face = "bold", hjust = 0.5),
    legend.position = "top",
    axis.title.x = element_text(family = "Georgia", face = "italic", size = 12),
    axis.title.y = element_text(family = "Georgia", face = "italic", size = 12),
    legend.text = element_text(family = "Georgia", face = "italic", size = 10)
) +
  scale_x_discrete(breaks = unique(bar_dat$year)) +
  
  scale_fill_manual(
    values = c("cat_a" = "#4E2A84", "cat_b" = "#836EAA", "cat_c" = "#E4E0EE"),
    labels = c("cat_a" = "Applications", "cat_b" = "Admitted students", "cat_c" = "Matriculants"),
               name = NULL)
  

rate_dat <- admin_dat %>%
  filter(year >= 2002) %>%
  select(year, contains("_rate")) %>%
  pivot_longer(
    cols = -year,
    names_to = "category", 
    values_to = "value"
  )

rate_labels <- admin_dat %>%
  filter(year >= 2002) %>%
  select(year, contains("_rate")) %>%
  pivot_longer(
    cols = -year,
    names_to = "category", 
    values_to = "value"
  ) %>%
  mutate(
    pct_label = str_c(value, '%'),
    value = case_when(
      category == "yield_rate" ~ value + 2, 
      category == "admission_rate" ~ value -2
    )
  )
  

plot_2 <- ggplot(data = rate_dat, aes(year, value)) +
  geom_line(aes(color = category)) +
  geom_point(data = rate_dat, 
             aes(color = category, shape = category),
             size = 2) +
  scale_shape_manual(values = c("yield_rate" = 8, "admission_rate" = 15),
                     labels = c("yield_rate" = "Yield Rate", "admission_rate" = "Admission Rate"),
                     name = NULL) +
  scale_color_manual(values = c("yield_rate" = "#4E2A84", "admission_rate" = "#E4E0EE"),
                     labels = c("yield_rate" = "Yield Rate", "admission_rate" = "Admission Rate"),
                     name = NULL) +
  theme_minimal() +
  geom_text(
    data = rate_labels,
    aes(label = pct_label),
    size = 2.5,
    color = "#716C6B",
    fontface = "bold"
  ) +
  labs(
    x = "Entering Year",
    y = "Rate",
    title = "Northwestern University \nUndergraduate Admissions 2001-2021") +
  theme(
    legend.position = 'top',
    panel.grid.minor.x = element_blank(),
    plot.title = element_text(family = "gothic_font", size = 20, face = "bold", hjust = 0.5),
    axis.title.x = element_text(family = "Georgia", face = "italic", size = 12),
    axis.title.y = element_text(family = "Georgia", face = "italic", size = 12),
    legend.text = element_text(family = "Georgia", face = "italic", size = 10)
  ) +
  scale_x_continuous(breaks = seq(2002, 2022, by = 1)) +
  scale_y_continuous(labels = function(x) paste0(x, "%"))

plot_1

plot_2

Challenge

Important

Challenge is optional for all students, but we recommend trying it out!

Using cdc_small dataset, re-create your own version inspired by the plot below.

Must haves:

• Use two non-standard fonts (one for labeling the point and the other for the axes)
• Use at least two colors (one for the added point, another for the rest of the points)
• A curved arrow used to label the point

Using Bilbo Baggins’ responses below to the CDC BRFSS questions, add Bilbo’s data point to a scatterplot of weight by height.

• genhlth - How would you rate your general health? fair
• exerany - Have you exercised in the past month? 1=yes
• hlthplan - Do you have some form of health coverage? 0=no
• smoke100 - Have you smoked at least 100 cigarettes in your life time? 1=yes
• height - height in inches: 46
• weight - weight in pounds: 120
• wtdesire - weight desired in pounds: 120
• age - in years: 45
• gender - m for males and f for females: m

Note

Adding non-standard fonts can be an adventure. I find this blog post to be extremely useful for adding fonts. Important packages for using non-standard fonts are showtext, extrafont, extrafontdb, and sysfonts. The last 3 generally just need to be installed (not loaded per session).

Hint:

• Create a new dataset (maybe call it bilbo or bilbo_baggins) using either data.frame() (base R - example in book) or tibble() (tidyverse - see help documentation for the function). Make sure to use variable names that exactly match cdc’s variable names. We have provided the tidyverse approach.
• Search google fonts to find some free fonts to use (can get free fonts from other locations)

Solution

# build dataset for Bilbo Baggins
bilbo <- tibble(
  genhlth  = "fair",
  exerany  = 1,
  hlthplan = 0,
  smoke100 = 1,
  height   = 46,
  weight   = 120,
  wtdesire = 120,
  age      = 45,
  gender   = "m"
)